CCOBRA Evaluation Results

Benchmark Information
Benchmarkfigural.json
Date2020-04-24 17:25
typeadaption
domainsspatial-relational
response_typesverify
data.test3ps.csv
data.train3ps.csv
data.train_person
corresponding_datatrue
Results
Prediction Accuracy
RandomModelLSTMAutoMLMLPMostFrequentAnswerPRISMRNNTransitiveClosureVerbal Reasoning00.20.40.60.8
Accuracy of Model PredictionsPredictive Accuracy

Predictive accuracies of the benchmark models, i.e., the percentages of correct predictions.

Subject-Based Boxplots
RandomModelLSTMAutoMLMLPMostFrequentAnswerPRISMRNNTransitiveClosureVerbal Reasoning0.40.60.81
RandomModelLSTMAutoMLMLPMostFrequentAnswerPRISMRNNTransitiveClosureVerbal ReasoningSubject Performance BoxplotPredictive Accuracy

This plot depicts boxplots for the models indicating individual subject performance. The data used for the plot are accuracies for individuals. Consequently, min and max refer to the accuracy of the worst and best matching subjects. The dots refer to the mean accuracies of individual participants.

Most-Frequent Answer Comparison

No data encodings available.